[AI Evaluation] Improve how chat history is displayed in reports

[shyamnamboodiripad]

• Replace the current prompt and response text boxes with a nicer chat bubble display that shows the entire conversation (including displaying the role (user v/s assistant v/s tool calls etc.) more naturally. [Fixed in Introduce Reason property on EvaluationMetric #6087 ]

• Make the Render Markdown button to be global to the entire report. [Fixed in Introduce Reason property on EvaluationMetric #6087 ]

• We could also introduce a Debugging View toggle to switch between a simple / friendly display format (where markdown rendering is turned on, tool call messages are hidden etc.) and a display format that is rawer (which could display additional details such as token counts / time stamps / latency (see [AI Evaluation] Include reporting for token counts and latency #5970) that are hidden by default).

FYI @peterwald

[shyamnamboodiripad]

The first two items in the above list were addressed via #6087. This issue now mainly tracks the Debugging view improvements above.

[shyamnamboodiripad]

We have made a bunch of improvements on the reporting side recently specifically focused around how the chat history and associated data (such as token counts) are visualized. The reporting functionality has also evolved quite a bit since this issue was originally logged - we have since added support for collapsible sections, filtering and metadata.

@peterwald I feel like we could wait on more user feedback before we work on the debugging view idea above. Closing this issue for now. We can always reactivate / create a new issue as needed based on future feedback. Let me know if you think otherwise.